Calibrating Profiles Between Roasters

Share |

Postby N3Roaster » Mon Jul 02, 2012 5:28 am

Recently I've been working with a couple different coffee roasters. These are very similar in design: same type of burners, similar air flow controls. The main difference is capacity (1Kg v. 12Kg). Both have a thermocouple measuring bean temperature which is consistent from one batch to the next, but the measurements at equivalent points of the roast were not the same between the two roasters. I suspect that this is a common issue with multi-roaster facilities and while it might not be applicable where the design of the roasters is less similar, I thought I'd share the solution that I came up with as a starting point for others in the same situation and perhaps other people here have solved the problem differently and might want to share their solutions.

I suspected that I would be able to get a good match between these roasters by observing the temperature at a few easily observable equivalent points (for example, first crack on these roasters always occurs at the same temperature though the measurements are different between the two roasters) and using these measurement pairs to define a mapping function based on cubic spline interpolation (the differences are slightly nonlinear throughout the roast). Before implementing this, however, I tried with five point pairs (color change from green to yellow, color change from yellow to light brown, start of first crack, start of second crack, and one point beyond this) using linear spline interpolation as I had already written most of the code needed to implement that and the results were so good that I do not believe it is worth the effort of implementing this with cubic splines. Having a good way of logging bean color throughout the roast would probably have made this calibration easier, though I do not have appropriate hardware for that so I had to rely on my eyes for three of the five points.

Linear spline interpolation is a simple numerical method. We take the equivalent measurement pairs described above and if we consider the measurements on one roaster to be our x coordinate and the measurements on the other to be our y coordinate, we can connect these points by line segments describable in the y = mx+b form you may remember from school. With this, it is possible to take a measurement from one roaster and obtain the equivalent measurement from the other. As such, the mapping can easily be done in a spreadsheet for those using manual logging.

The attached graph shows on the darker blue line measurements from the thermocouple of the smaller roaster with a lighter blue line showing the equivalent temperature on the larger roaster according to this method. The specifics of the mapping will, of course, vary from one pair of machines to the next.

heating.png
heating.png (13.78 KiB) Viewed 9482 times
Neal
N3Roaster
 
Posts: 64
Joined: Tue Mar 16, 2010 4:30 pm
First Name: Neal
Last Name: Wilson
City: Racine
State: WI
Zip Code: 53405
Company: Wilson's Coffee & Tea
Occupation: Roastmaster
Roasting Since (Year): 2000
Website: https://typica.us
Location: Racine, WI
Twitter: N3Roaster
YouTube: N3Roaster

Postby N3Roaster » Tue Dec 18, 2012 4:59 am

and here's a video that goes over the technique and describes how to set that up in software now that I've made that feature more generally available. I'm still amazed that this works at all, to be honest, but I've been doing this for several months now developing roast profiles for many coffees that are very different from each other and demand very different roasts and it keeps working so I have to assume that at least for the equipment that I'm using this is a valid technique.
Neal
N3Roaster
 
Posts: 64
Joined: Tue Mar 16, 2010 4:30 pm
First Name: Neal
Last Name: Wilson
City: Racine
State: WI
Zip Code: 53405
Company: Wilson's Coffee & Tea
Occupation: Roastmaster
Roasting Since (Year): 2000
Website: https://typica.us
Location: Racine, WI
Twitter: N3Roaster
YouTube: N3Roaster

Postby ChrisSchooley » Wed Jan 23, 2013 11:39 pm

Neal, as always, this is so boss. Thanks for sharing. This is extremely useful data as well for a number of roasters, specifically thinking about sample roasting to production and comparing those profiles.
ChrisSchooley
 
Posts: 174
Joined: Mon Mar 01, 2010 3:35 pm
First Name: Christopher
Last Name: Schooley
City: Fort Collins
State: CO
Zip Code: 80526
Company: Coffee Shrub
Occupation: Roaster, Micro-Seller of Green Coffee
Roasting Since (Year): 2001
Website: http://www.coffeeshrub.com/.
Location: Fort Collins, CO.

Postby Phil.Johnson » Sun Jan 27, 2013 7:20 am

N3Roaster wrote:and here's a video that goes over the technique and describes how to set that up ...


Awesome work on the video. Well edited and well put together. Succinctly states the problem and takes you through to the solution very straightforwardly.

It would be interesting to see if such mapping would work with roasters from different manufacturers -- or for different style roasters -- like sample roasters that measure the air temperature versus a roster that measures the temperature of the bean bed.

If anything, it might point out what variables really matter versus variables that have "a lot of play".
Phil.Johnson
 
Posts: 29
Joined: Thu Aug 25, 2011 5:48 am
First Name: Phil
Last Name: Johnson
City: Sheldonville
State: MA
Zip Code: 02070
Company: Sheldonville Roasters
Occupation: Roaster/Owner
Roasting Since (Year): 2003


Return to “%s” Profiles & Heat Transfer

Who is online

Users browsing this forum: No registered users and 1 guest